Privacy Preserving Protocol for Two-Party Classifier Over Vertically Partitioned Dataset Using ANN

Smitha Iddalgave, Sumana M

Abstract: With the emergence of distributed computing privacy preservation has become a priority concern. Privacy in the field of data mining can be ensured by having secure computations. Data mining in distributed scenario deals with data from multiple data providers. The providers have to be assured about the safety of their data. Hence; rather than having a trusted party (network) which can collect data from providers and perform meaning classification on the combined data; we propose a two party classifier which allows the network participants to work on their dataset and communicate with each other in a secure manner using encryption schemes to establish relation between their data without revealing any of their private data to each other. The participants can only learn about their input and output values. The protocol is implemented on vertically partitioned dataset; as with horizontal partitioning the processing becomes sequential with the output of one network being fed to other for further processing. This protocol can be extended for multiple participants and checked for its privacy-preservation property.

Keywords: Neural network, backpropagation, learning, privacy-preserving, cryptographic scheme

How to Cite?: Smitha Iddalgave, Sumana M, "Privacy Preserving Protocol for Two-Party Classifier Over Vertically Partitioned Dataset Using ANN", Volume 3 Issue 6, June 2014, International Journal of Science and Research (IJSR), Pages: 1654-1660, https://www.ijsr.net/getabstract.php?paperid=2014581, DOI: https://dx.doi.org/10.21275/2014581